Figure 3.

Polygenic covariance structure between males and females

(A) Our analysis of the polygenic covariance between males and females is based on sex-stratified GWASs. We modeled the sex-stratified GWAS estimates as sampled with error from true effects arising from a mixture of possible covariance relationships between female and male genetic effects. As an example, shown are illustrations for three possible relationships of the same qualitative nature—perfectly correlated effects that are also larger in females—and the mixture weights estimated for each in the case of diastolic blood pressure.

(B–F) Each box shows the sum of weights placed on all covariance relationships of the same qualitative nature, as specified by relative magnitude (horizontal axis) and correlation (vertical axis) between male and female effects. The full set of pre-specified covariance matrices is shown in Figure S2, and the weights placed on each of them for each trait are shown in Data S1–27. All weights shown are percentages of non-null weights; i.e., the weight divided by the sum of all weights except for the one corresponding to no effect in either sex.